Introduction
In 2020 together with a fellow volunteer made measurements on the 33 flowering plants
at Warburg and using discriminant analysis was able to separate them into 3
groups, Epipactis leptochila (El), Epipactis helleborine (Eh) and what I
thought might be hybrids (Ex). The work
was described in Swainbank and Boseley, (British and Irish Botany Vol 3 (No2) 2021
A leaf sample of one of those putative hybrids, Plant #35,
was sent for molecular analysis at the Royal Botanic Gardens Edinburgh.
In 2021, I and three others repeated the exercise at Warburg when there were just 10 flowering plants, and
combined this new data with that from 2020. The discriminant analyses was then run again.
In addition I made measurements for the same variables on a
number of plants from 3 other locations, Pulpit Hill in Buckinghamshire ,
Buckle Wood in Gloucestershire, and Garth Wood in Glamorgan to make a
comparison. Pulpit Hill is close to
Whiteleaf where the species identification of helleborines has always been
controversial. I would have liked to
have included plants from Whiteleaf in the comparison, but in 2021 I could find no extant
plants. The plants at Garth Wood were
the subject of a study by Lewis et al suggesting that some should be described as E. leptochila var cordata.
A subset of the combined dataset is shown in Appendix 1.
I used discriminant analysis to compare and contrast the
plants from these 4 locations.
Molecular analysis on a putative
hybrid from Warburg
I confidently expected
that Plant #35 at Warburg would be a hybrid between E. helleborine and E.
leptochila, i.e. E. x stephensonii. Molecular
analysis was carried out on a leaf sample from the suspected
hybrid (#35) and two reference plants clearly identified as E.
helleborine (plant #26) and E. leptochila (plant #38) from
Warburg. Using these two reference samples and data from
Genebank (8 sequences of E. helleborine and 4 sequences from E. leptochila) 3 base
pairs (SNPs) were identified in the nuclear ITS region which are
species-specific to the parental species, i. e. only occur in one species but
not the other. The 3 SNPs in the suspected hybrid matched the
species-specific SNPs of E. leptochila suggesting that plant #35 is
E. leptochila and not a hybrid. The work was undertaken by Dr Markus
Ruhsam for which I am hugely grateful. Why the morphometric
measurements should suggest misleadingly that there is a difference is not
obvious..
Morphology of the Warburg plants
There were 4 plants that I thought might be hybrids in 2020,
before I received the results of the molecular analysis on one of them. A further two plants that flowered in 2021
had similar characteristics, whilst one of the 2020 plants flowered again. A photo of a single flower from each is shown
in Fig 1, although actually in the case of plants 35, 40, 78 and 139 I took
photos of most of the flowers on the spike and some flowers were photographed
on several occasions to demonstrate their appearance on ageing.
Fig 1 Warburg Plants Labelled as Putative Hybrids
Plant #102 Plant #35 (Fresh) Plant #40 Plant #75
Plant #78 Flower 1 Plant 139 Flower #2
Discriminant analysis on the Warburg plants
I repeated the discriminant analysis I undertook in 2020,
but using the combined data from 2020 and 2021, to separate the plants which I
gave a priori identifications as El (E. leptochila) Eh (E. helleborine) and Ex (my putative hybrids or perhaps puzzles). Those a
priori identifications were influenced by the lip shape, colour of pedicel
attachment to the stem, leaf sizes and arrangement.
I ignored any possible annual variation. The variables that I measured were:
- ·
Epichile length and width
- ·
Hypochile width
- ·
Viscidium width (0 if not present)
- ·
Lateral sepal length and width, visible from the
front
- ·
Lateral petal length and width, visible from the
front
- ·
Dorsal sepal width, visible from the front
- ·
Leaf red, green and blue components, after
standardising to a calibrated grey scale
In 2021 I and 3 volunteers made measurements on several flowers from each
plant and averaged the results.
Vegetative variables measured, but not used in the
discriminant analysis other than to derive an algorithm to make predictions of the
identity of non-flowering plants were:
- ·
the length and width of each of the bottom 3
leaves,
- ·
the clockwise leaf turn from one to the next of each
of the bottom 4 leaves.
I had data for 44 plants but took out 7 plants because the
data was incomplete.
The discriminant analysis separated the plants cleanly to
their a priori groups. (Fig 2) A histogram of the first discriminant
component which described most of the variability is shown in Fig 3.
Fig 2 The two
discriminant functions derived from perianth and leaf colour variables plotted
against each other
Fig 3 A
histogram of the principal discriminant component derived from perianth and
leaf colour variables
In Fig 3 two points are worth noting. Firstly the plants
grouped as Ex sit midway between the Eh and El groups suggesting that there is some
form of hybridisation or introgression
going on. Secondly, in it can be seen
that Plant #35, which molecular analysis showed to be E. leptochila rather than a hybrid, is the closest of the Ex plants
to the El group. It had a wider epichile
and a more open, wider flower than was the average for El.
The variables which most influenced the separation of El
from Ex were the epichile width, and the size of the viscidium if present,
whilst separation of El from Eh was mostly influenced by those same two
variables but also by the epichile length.
Epichile length is to be expected because this was an underpin of the a priori identifications, though if this
was omitted from the analysis the weighting of other factors still gave clear
separation.
I ran a discriminant analysis on the vegetative
characteristics of the flowering plants, where any missing data was replaced by
the median for that taxa (obviously giving some improvement to separation), the
variables being
·
Length to width ratio of each of the lowest 3
leaves
·
Clockwise leaf turn, one to the next of the
bottom forur leaves
·
Ratios of the colour components, red to green,
red to blie and green to blue , with those colours standardised to a grey
scale.
The discriminant achieved 81% accuracy in separating plants
to their a priori groups. One Eh reported as Ex, 3 of the 16 El
reported as Eh, but the 6 putative hybrids were less well classified with one
reporting as Eh, and two as El.
Intriguingly the contentious plant #35 stayed loyal to the Ex
group. A plot of the two discriminant
functions follows in Fig 4
Fig 4 The two
discriminant functions derived from leaf colour, size and arrangement
Whilst leaf colour had by far the biggest influence on separation,
on its own it only achieved a separation accuracy of 63%, so leaf turn and leaf
size cannot be ignored.
Comparing Warburg plants with plants
from 3 other locations
I grouped the plants at Warburg which I concluded were
either E. leptochila or putative
hybrids and compared them with plants from 3 other locations which based on
observation were all E. leptochila: 6
plants from Buckle Wood in Gloucestershire, (an example in Fig 5) 2 plants from
Pulpit Hill in Buckinghamshire (both plants in Fig 6) and 12 plants from Garth
Wood in Glamorgan (2 examples in Fig 7)
Fig 5 Plant #10 Buckle Wood
Fig 6 Pulpit Hill Plants P1 and P2
Fig 7 Garth Wood Plants #7 (var cordata) and #8
I made the same measurements on morphological variables were
made on all these plants as those at Warburg.
Running a discriminant analysis showed that location could differentiate
the plants and a plot of the 2 principal components against each other shows a
clear differentiation between plants (Fig 8)
Fig 8 The two
discriminant functions derived stature, perianth dimensions and leaf colour at
4 locations
The assignment accuracy was 100%.
Buckle Wood plants look much more like the classic E. leptochila s.s than those from the
other locations; indeed this was borne out when the impact each variable had on
separation efficiency.
Plants from Garth Wood differed from those at Buckle Wood
mainly by having a wider epichile and hypochile, and they were taller. Ignoring plant height as a variable, as an
obvious cause of possible sampling bias, the separation still held.
Warburg plants again had wider hypochiles than those at
Buckle Wood, but the lateral sepals were also wider. Garth Wood plants had wider epichiles and a
stronger blue leaf colour component than those at Warburg.
Given the relative closeness of Warburg and Pulpit Hill, I
had expected the plants at Warburg taken in aggregate to be indistinguishable
from those at Pulpit Hill, though I only had a couple of plants from the latter
for comparison. They were different
however as a group; Warburg plants had
wider hypochiles, less likelihood of a well-developed viscidium and the leaves
had a larger green component in the leaf.
But the two Pulpit Hill plants differed from each other, one being close
to the Warburg plants, the other more akin to Buckle Wood plants despite
superficially being similar to each other (Fig 6).
Again a surprise was the result from a further discriminant
analysis comparing the Pulpit Hill plants with the 6 putative hybrids. The 2
Pulpit Hill plants were more similar to each other than to the 6 putative
hybrids from Warburg, ( including of course Plant #35 given by molecular
analysis as E. leptochila). The histogram for the principal component is
shown in Fig 9
Fig 9 A
histogram of the principal discriminant component derived from perianth and
leaf colour variables comparing Warburg putative hybrids and Pulpit Hill plants
In fact though, looking at the individual plants, Warburg
Plant #75 and Pulpit Hill Plant #P2 are closer to each other than to their
siblings, illustrated for example by the following plot of epichile width and hypochile
width (Fig 10). If any of the plants are
possible hybrids, then these two are the most likely candidates. Leaf samples of Plant#75 at Warburg have
been collected for molecular analysis.
Fig 10 Hypochile width v epichile
width for Warburg’s putative hybrids and Pulpit Hill plants
Garth WoodOf the 12 plants from Garth Wood, 4
fell within the description of var.
cordata, (Co) and a further two were almost cleistogamous,(Cl) with the flower barely opening. I could separate these two groups from the
rest by discriminant analysis (Fig 11).
As might expected the key determinants were the epichile width and
length and the hypochile width: Fig 12 is a plot of epichile width to
hypochile width showing that the width of the epichile is much more
pronounced in var. cordata plants. They are clearly different from
the other Garth Wood plants.
Fig 11 The
two discriminant functions from an analysis of the Garth Wood plants (Cl =
cleistogamous, SS = sui strict, Co = variety cordata)
Fig 12 Hypochile width v epichile
width for Garth Wood plants
Conclusions
1 Some plants at Warburg can be differentiated from either E. leptochila or E. helleborine, replicating a result from 2020 with more data, yet
molecular analysis showed that one of those plants was not a hybrid. Perhaps there is a degree of introgression
and to examine this further samples have been taken from 5 more plants for
molecular analysis.
2 E. leptochila is
variable, perhaps because it is autogamous such that geographically separate
populations are distinct and dissimilar from plants at other locations.
3 The description of some plants at Garth Wood as a variety
(E. leptochila var cordata) appears
justified.
Appendix 1 Perianth
dimensions, leaf colour and stature for E. leptochila plants and putative
hybrids from Warburg and 3 other locations
